RDP 2016-04: Housing Prices, Mortgage Interest Rates and the Rising Share of Capital Income in the United States 1. Introduction

I believe that the right model to think about rising capital-income ratios and capital shares in recent decades is a multi-sector model of capital accumulation, with substantial movements in relative prices … [i]ndeed, large upward or downward movements of real estate prices play an important role in the evolution of aggregate capital values during recent decades Piketty 2016

Piketty (2014) documents how the share of aggregate income going to capital in the United States (and other advanced economies) followed a U-shaped pattern in the post-war era; it fell between the 1940s and 1970s but has risen since then.^[1] Rognlie (2015) has subsequently shown that much of the rise in the net capital income share in the post-war era is due to the housing sector (Figure 1). The share of total income going to the owners of housing capital (or ‘rental income’) in the United States gradually rose from around 3 per cent in 1950 to 7 per cent in 2014.

The long-run rise in the share of spending on housing in the US economy is not specific to the national accounts, but can be observed across a range of household surveys, including the American Housing Survey, the Census and the Consumer Expenditure Survey (Albouy, Ehrlich and Liu 2014). The secular rise in the ‘housing capital share’ of the economy is also not specific to the United States but has occurred in almost every advanced economy over the past three decades (Rognlie 2015). The broad-based nature of the secular rise in the housing capital share – both across surveys and across countries – suggests it is not a measurement artefact but a genuine phenomenon.

The aim of this paper is to examine why the housing capital income share rose in the United States over recent decades. Structural factors such as an increase in the home ownership rate and an increase in the average size and quality of housing are important in explaining the increase in the housing capital income share in the period immediately after the Second World War. However, these structural factors appear to have been less important in explaining the ‘rise of housing’ in the period since the early 1980s.

Several research papers (e.g. Rognlie 2015; Bonnet et al 2014; Weil 2015), print articles (e.g. The Economist 2015) and blogs (e.g. Smith 2015) have hypothesised that the secular increase in the housing share of the economy over this time might be due to some combination of lower interest rates, higher mortgage debt and constraints on home building (due to either geographic constraints or land zoning restrictions). But, to the best of my knowledge, no previous study has empirically documented the links between the trend increase in the share of housing capital income on the one hand, and financial market liberalisation, mortgage interest rates and housing supply constraints on the other.

In theory, the long-run rise in the housing share of the economy is somewhat puzzling. To the extent that the consumption of housing services is a necessity, housing demand should be income inelastic; as households get richer, they should spend less on housing services. Similarly, as an economy grows, the share of aggregate spending on housing should fall, not rise. Classical studies typically pointed to evidence that the income elasticity of housing is less than unity, consistent with housing being a necessary good. But later research has pointed to more mixed evidence, with some studies finding an elasticity well above unity (Albouy et al 2014).

This empirical puzzle can be reconciled with theory by noting that a home consists of both a land component and a structure component; the building structure is a necessary good but the land is, quite literally, a ‘positional’ or luxury good (Frank 2005). So, to the extent that land is a luxury good, we might expect an increase in income to be associated with higher demand for housing services. Moreover, if there are very few substitutes for housing, and hence demand is price inelastic, then rising housing prices could cause the housing expenditure share to rise, even as income rises (Albouy et al 2014).

Along these lines, I reconcile the theory with the facts by appealing to the insensitivity of housing supply to changes in the relative price of housing in some large cities of the United States. Specifically, I argue that consumer price disinflation and the deregulation of the US mortgage market during the 1980s and 1990s acted as positive credit supply shocks (with high inflation and credit market regulation in the 1970s acting as artificial borrowing constraints). The subsequent decline in nominal interest rates lowered the cost of owning and so effectively increased the demand for housing for credit-constrained households (Ellis 2005). The resulting increase in housing demand led to higher relative prices for land in areas that are constrained in terms of new housing supply. The rise in the relative price of land, in turn, led to an increase in the (nominal) share of spending on housing. Given that housing supply constraints are typically most prevalent in the largest US cities, they contributed disproportionately to total spending on housing (and income accruing to the owners of housing) in the overall economy. While this paper focuses on the empirical evidence, Sommer, Sullivan and Verbrugge (2013) and Stiglitz (2015) outline theories that are consistent with this hypothesis.^[2]

First, I document some new stylised facts about housing income and spending in the United States. I follow Piketty and Zucman (2014) and Rognlie (2015) in undertaking a detailed examination of the national accounts, but I take the further step of decomposing the data by geographic region (e.g. states and metropolitan areas) and also by different types of housing (e.g. owner-occupied and tenant-occupied). Second, I explore the determinants of the secular rise in the housing share of the economy by exploiting both cross-sectional and time-series variation in factors such as housing prices, interest rates and land supply constraints (as documented, for example, by Saiz (2010)). In a state-level panel regression framework, I test the following hypotheses:

• H1: Lower nominal interest rates are associated with higher net housing capital income (as a share of total income) across US states and over time:
  • H1a: Lower real interest rates are associated with higher net housing capital income (as a share of total income)
  • H1b: Lower consumer price inflation is associated with higher net housing capital income (as a share of total income)
• H2: The negative correlations in H1 will be strongest in US states that are constrained by housing supply.

My main findings are as follows:

1. The rise in the share of housing capital income is due to an increasing share of imputed rent going to home owners (owner-occupied property) rather than an increasing share of market rent paid to landlords (tenant-occupied property).
2. The rise in the share of housing capital income is due to an increase in the relative price of housing and is fully concentrated in states that are estimated to be constrained by the supply of new housing.
3. The rise in the share of housing capital income is associated with long-run declines in both real interest rates and inflation, with these effects being particularly strong in supply-constrained states.

Ultimately, I argue that the rise in the share of housing capital income can be traced to an aggregate demand shock (the expansion of credit brought about by financial deregulation and disinflation), along with constraints on the supply of new homes in some large US cities.

I focus specifically on the US experience as US state-level housing markets are likely to be similar in both observable and unobservable characteristics (or at least more similar than housing markets in different countries). This limits the impact of any confounding factors and helps to pin down the causal effect of changes in interest rates and housing supply constraints on housing capital income.

Nevertheless, I suspect a similar story of financial deregulation, disinflation and housing supply constraints might explain the patterns observed in other advanced economies too, especially considering the similar timing to the United States of financial deregulation and disinflation, as well as the concentration of populations in the largest cities.

This paper is related to several strands of the literature. First, there is a large and expanding literature on the determinants of wealth and income inequality (e.g. Piketty and Zucman 2014). I will not touch directly on the issue of inequality in this paper, although the analysis reveals some interesting subtleties about the recent increase in the concentration of wealth amongst land owners; it is not landlords per se but home owners that have been ‘winning the battle’ over wealth shares, aided by lower interest rates. Second, by examining the links between financial deregulation, disinflation and housing capital income, I closely follow the extensive literature that examines the causal effect of credit supply shocks on housing prices (e.g. Ellis 2006; Favilukis, Ludvigson and Van Nieuwerburgh 2010; Duca, Muellbauer and Murphy 2011; Favara and Imbs 2015). Third, I highlight the role of housing supply constraints in driving the long-run trend increase in housing spending and hence touch on a large literature in urban economics that examines the links between land supply, housing prices and rents (e.g. Gallin 2003; Saiz 2010; Gyourko, Mayer and Sinai 2013; Hilber and Vermeulen 2016).

Footnotes

The flipside of this has been a decline in the labour share of the economy (e.g. Guscina 2006; Ellis and Smith 2010; Elsby, Hobijn and Şahin 2013; Karabarbounis and Neiman 2014). [1]

Borri and Reichlin (2015) suggest an alternative explanation for the secular rise of housing. Based on a two-sector, life-cycle model, they suggest that if productivity in the manufacturing sector grows more rapidly than that in the housing construction sector, then this will lead to a higher relative price for housing services in equilibrium. And if housing demand is sufficiently price inelastic, this in turn will contribute to a higher nominal share of spending on housing. I provide cross-sectional evidence that the secular rise of housing has been fully concentrated in states that are constrained by available land supply, whereas the ‘housing cost disease’ hypothesis of Borri and Reichlin (2015) would predict that it has occurred mainly in states experiencing fast manufacturing productivity growth. Moreover, the increase in the relative price of housing observed in most advanced economies occurred at a time when productivity growth in domestic manufacturing was slowing, which seems to argue against this version of events. In contrast, Davis and Ortalo-Magné (2011) outline a model that predicts that variation across cities in the relative price of housing is independent of housing supply conditions and depends purely on differences in average incomes. My results also contradict the predictions of that model. [2]